InstantX
InstantX
We don't have a specific plan to release this part of codes by now, but we do encourage the community to implement and PR.
Added.
Check this [node](https://github.com/cubiq/ComfyUI_IPAdapter_plus).
In ComfyUI, it call style / composition transfer.
But SD2.0 is not a common adopted model, why do you need it?
@KovenYu We haven't tested on inpainting model yet, although they have the same structure except for the first input channel. Are you using [stable-diffusion-2-inpainting](https://huggingface.co/stabilityai/stable-diffusion-2-inpainting)? Or [stabilityai/stable-diffusion-2](https://huggingface.co/stabilityai/stable-diffusion-2)? I believe they both...
We will check it today.
For [SDXL-Inpainting](https://huggingface.co/diffusers/stable-diffusion-xl-1.0-inpainting-0.1), it works as below. But as IP-Adapter haven't released checkpoint for SD2.0, we need to retrain it. The risk here is that we find the SD1.5 has less...
It is a connection problem, where you cannot download directly from hf, use hf-mirror and download all models manually would help. Below is an example how to use. ``` export...
You should use `torch.float32`